Complex Capital Buffers Coincided With More Headroom and Slower Corporate Lending

An increase in the complexity of euro-area capital buffers was associated with 26 to 36 basis points more capital headroom and roughly 0.5 to 0.9 percentage points less quarterly growth in existing corporate loans. The estimates cover 1,411 banking-union institutions. They come from an ECB working paper published in September 2026, not an adopted policy finding.

The distinction matters. The paper tests the form of capital regulation after controlling for its combined level. Its question is whether the number, geographic spread and reset frequency of overlapping buffers affect bank behaviour even when headline requirements look similar.

The answer is an association, not a welfare verdict. More tailored buffers can target risk more precisely. They can also make future requirements harder for banks to forecast. The paper finds evidence consistent with banks paying for that uncertainty through extra capital and slower growth in existing bank-firm loan exposures.

One capital ratio can hide a complicated stack

The authors build a quarterly Macroprudential Fragmentation Index from three components. Breadth counts positive-rate countercyclical, systemic-risk and systemically important bank buffers applying at the same time. Churn counts upward rate changes over the previous four quarters. Dispersion uses an inverted concentration index to measure how much exposure-based requirements depend on decisions across several countries.

Each component is standardised. The index takes their equal-weighted average. The construction window runs from the first quarter of 2015 through the fourth quarter of 2024.

That raw measure rises with things that can independently affect lending. Bigger banks operate across more jurisdictions. Higher combined buffer requirements often involve more instruments. Buffers also change with the financial cycle. The authors therefore regress the raw index on the combined buffer requirement, bank assets and the ECB's composite financial-cycle indicator, then use the residual in their analysis.

This is useful statistical separation. It is not proof that authorities assigned complex policies at random. The residual can still co-move with unobserved bank or policy conditions. Alternative index weights and residualisation methods produce similar estimates, which reduces sensitivity to the authors' chosen formula without removing that identification limit.

Banks kept more room above the line

The headroom analysis covers quarterly data from the first quarter of 2021 through the fourth quarter of 2024. Headroom is the bank's common-equity Tier 1 ratio minus its overall capital requirement and Pillar 2 Guidance.

In the main headroom table, a one-standard-deviation increase in the adjusted index is associated with 26 basis points more headroom in the preferred fixed-effects specification. Less saturated models produce about 35 basis points. Alternative residualisation procedures produce 35 to 36 basis points, which explains the working paper abstract's roughly 36-basis-point summary.

That range supports a precautionary-retention interpretation. It does not show that complexity caused every extra euro of capital. It does show that the result is not a mechanical restatement of a higher combined requirement.

The lending test matches the index to AnaCredit exposures above EUR 25,000. It compares banks lending to the same firm in the same quarter, which absorbs firm-level demand shocks shared across those lenders. Across almost 2.95 million bank-firm-quarter observations, the estimated reduction in existing-relationship loan growth narrows from 0.88 to 0.51 percentage points as the models add stronger bank and country controls. One interest-rate specification produces a 0.96-point estimate.

That narrowing is informative. The larger estimates use both differences between banks and changes over time. The preferred models add bank effects, so they ask whether lending growth slowed as the same bank's exposure to fragmentation increased. Persistent differences in business model, ownership and risk appetite then drop out. The remaining estimate is smaller, but it is more closely aligned with the paper's claim about changes in the regulatory environment.

Capital position changes the relationship. In the paper's interaction model, the estimate for a bank with no headroom is a 0.668-percentage-point reduction in loan growth. Each additional percentage point of headroom offsets 0.055 points of that association. Banks closest to the supervisory threshold therefore show the strongest response.

The paper's mechanism is plausible but indirect. More complex buffers may raise the chance that a bank misjudges its future requirement, making extra capital valuable insurance. Yet higher headroom is not automatically an economic cost. It can absorb losses and support lending in a downturn. The measured cost in this study is the slower growth of incumbent loan relationships, not the capital cushion itself.

The paper identifies a bank-by-bank regulatory-form association. It does not establish a fall in total corporate finance.

The missing margins decide the policy case

The outcome is growth in positive, existing bank-firm exposures. The paper does not estimate whether firms formed new lending relationships, whether other banks replaced the slower growth, or whether bond and non-bank finance filled the gap. It observes interest rates and original maturity, but uses them as controls rather than testing pricing or maturity as outcomes.

The sample design reinforces that boundary. The regressions retain positive outstanding amounts with reported annualised rates and identify supply through firms borrowing from multiple banks. A loan that was never originated, a relationship that ended at zero, or a borrower that moved outside the banking sample requires a different outcome design. Firm-by-quarter controls solve a demanding credit-demand problem inside the observed relationships. They do not turn those relationships into the firm's full financing account.

Those omissions prevent a broader financing-contraction claim. They are especially important because the same authors' 2024 work on the level of post-pandemic buffer increases tested new relationships and firm-level substitution. It found that substitution mitigated real effects for that different policy treatment. The result cannot be imported into the complexity paper. It shows why the missing tests could change the conclusion.

The paper also does not measure whether more precise buffers reduce losses, defaults or crisis costs. A regulator could rationally accept modest operating friction if tailored requirements materially improve resilience. Conversely, the case for simplification strengthens if the same protection can be delivered with fewer overlapping instruments and more predictable resets. The ECB has already said there is scope to reduce unwarranted complexity without weakening resilience.

The immediate policy test is narrow. Preserve the risk coverage, simplify the stack, and then measure new credit, pricing, maturity and substitution alongside existing-loan growth. Until those outcomes and loss benefits are observed together, the paper is evidence that regulatory form matters. It is not evidence that simpler rules are automatically safer or that complex buffers reduced total corporate funding.